Contribution of temporal variation for price discovery in high-dimensional arbitrage portfolios

• Grant number: 14/26448-0
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: July 01, 2015
• End date: March 31, 2017
• Field of knowledge: Applied Social Sciences - Economics - Quantitative Methods Applied to Economics
• Principal Investigator: Pedro Luiz Valls Pereira
• Grantee: Marilia Gabriela Elias da Silva
• Host Institution: Escola de Economia de São Paulo (EESP). Fundação Getúlio Vargas (FGV). São Paulo , SP, Brazil
• Associated research grant: 13/22930-0 - Price discovery in high-dimensional arbitrage portfolios, AP.TEM

Abstract

This research project aims to extend the standard methodology for price discovery to tackle time-varying information shares. The existent information share measures are constant over time because it assumes a stable contemporaneous correlation between market. However, recent empirical results based on recursive estimates and sample splitting indicate that the contribution to the price discovery may vary across market cycles. This project for a post-doctorate aims to tackle this issue by contriving a price discovery methodology that allows for stochastic covariance matrices and for irregularly-spaced-in-time data. The advantage of employing tick data is that trade duration, as measured by the time interval between transactions, reveals relevant information about how timely the different asset prices in the arbitrage portfolio react to news and hence about the price discoveryfprocess. Such a setting yields information shares that depend on market intensity, as measuredby the trade durations. As for allowing for time-varying covariance matrices, the post-doc studentfwill have to extend the existing methods to three setups. In the first, the covariance matrix is stochastic and part of the difusive component of a continuous-time multivariate diffusion process. The plan is to estimate the integrated covariance matrix over a fixed time interval (say, a day or week) using realized measures of covariation. In the second, the covariance matrix is constant only within regimes, with a latent state variable driving the determination of the latter by means of a Markov transition matrix. Finally, the third extension attempts to tackle the dependence of the price discovery mechanism on market cycles in a more general setting in which not only the covariance matrix, but also the cointegrating space of the price system do change across regimes.